Confidence Interval for a Proportion

We want to estimate a population proportion p (binomial probability of success).

Point Estimate + Margin of Error (Margin of Error is also called Error Bound)
x number of successes in sample
Point Estimate: Sample Proportion: p' = =
n total number in sample
p' q '
Error Bound: EBP = (critical value)(standard error) = Zα
2 n
The critical value depends on the confidence level. Zα is the Z value that will put to an area equal to the
2
confidence level (CL) in the middle of standard normal distribution N(0,1)
Zα tells us how many "appropriate standard deviations" to enclose about the point estimate,
2
where the "standard error" p' q ' is the appropriate standard deviation for a proportion
n
p' q '
Confidence Interval: p′ + EBP which is p′ + Zα
2 n

Confidence Interval for a Mean µ when σ is known

We want to estimate the population average µ and we already know the population standard deviation σ .
Point Estimate + Margin of Error (Margin of Error is also called Error Bound)
Point Estimate: Sample Average (Sample Mean) x
 σ 
Error Bound: EBM = (critical value)(standard error) = Zα  
2  n
The critical value depends on the confidence level. Zα is the Z value that will put to an area equal to the
2
confidence level (CL) in the middle of standard normal distribution N(0,1)
Zα tells us how many "appropriate standard deviations" to enclose about the point estimate,
where the "standard error" σ is the appropriate standard deviation for the sample mean
2

n
 σ 
Confidence Interval: x + EBM which is x + Zα  
2  n

Confidence Interval for a Mean µ when σ is NOT known

We want to estimate the population average µ and we do not know the population standard deviation σ.
We use the sample standard deviation s to estimate the population standard deviation σ
Point Estimate + Margin of Error (Margin of Error is also called Error Bound)
Point Estimate: Sample Average (Sample Mean) x
 s 
Error Bound: EBM = (critical value)(standard error) = t α  
2  n
The critical value depends on the confidence level. t α is the t value that will put to an area equal to the
2
confidence level (CL) in the middle of the student t -distribution with n − 1 degrees of freedom
t α tells us how many "appropriate standard deviations" we need to move away from the point estimate,
2
where s is an estimate of the standard error ("appropriate standard deviation") for the sample mean
n

Confidence Interval: x + EBM which is x + t α  s 

2  n

YOU ARE NOT PERMITTED TO BRING A PRINTOUT OF THIS PAGE AS YOUR NOTES FOR AN EXAM OR QUIZ.
Page 1 of 8
You CAN write whatever information you want from this page into your handwritten notes for exams or quizzes.
Interpreting the Confidence Interval:
For a proportion: 2 ways to word it
We estimate with _____% confidence that the true proportion of the population that
describe the population parameter in the situation of this problem is between ______ and ________
We estimate with _____% confidence that between _______% and _______% of the population
describe the population parameter in the situation of this problem

For a mean (average)

We estimate with _____% confidence that the true population average (or mean) describe the population
parameter in the situation of this problem is between _______ and _________

Finding the Point Estimate and Error Bound (Margin of Error) if we know the Confidence Interval:
The interval is (lower bound, upper bound)
Point Estimate = (lower bound + upper bound)/2 Error Bound = (upper bound – lower bound)/2

What does it mean to be CL% confident?

If we were to take repeated samples and calculate many confidence interval estimates based on those samples,
then we would expect that CL% of the confidence interval estimates would be “good estimates” that would
enclose (capture) the true value of the population parameter we are trying to estimate.
If we were to take repeated samples and calculate many confidence interval estimates based on those samples,
then we would expect that 100% − CL% of the confidence interval estimates would be “bad estimates” that
would NOT enclose (capture) the true value of the population parameter we are trying to estimate.

Note that the confidence interval is about proportions or averages. It is not about individual data values.
It does not mean that CL% of the data lies within the confidence interval.

To find Z that puts the area equal to the confidence level “in the middle”
CL tells use the area in the middle α α
α = 1 − CL is “outside”, split equally between both tails 2 2
α is in one tail.
2
− Zα 0 Zα
To find Zα : invnorm(1− α , 0, 1) 2 2
2 2
OR you can use invnorm( α , 0, 1) and take absolute value (drop the "−" sign)
2
Without calculator: Use a standard normal probability table to find Z.

To find t that puts the area equal to the confidence level “in the middle”
CL tells use the area in the middle α α
α = 1 − CL is “outside”, split equally between both tails 2 2
α is in one tail. df = degrees of freedom = n−1
2
− tα 0 tα
To find tα : TI-84+: invT(1− α , df) 2 2
2 2
OR you can use invT ( α , df) and take absolute value (drop the "−" sign)
2
TI-83,83+: Use INVT program; ask instructor to download it to your calculator: PRGM INVT
enter area to the left and df after prompts: area to left is 1− α ; (if using α as area to left , drop the "−" sign)
2 2
Without calculator or if calculator does not have inverse t: Use a student’s-t distribution probability table.
Page 2 of 8
Value of t is found at the intersection of the column for the confidence level and row for degrees of freedom
YOU ARE NOT PERMITTED TO BRING A PRINTOUT OF THIS PAGE AS YOUR NOTES FOR AN EXAM OR QUIZ.
EXAMPLE 1: CONFIDENCE INTERVAL ESTIMATE for an
unknown POPULATION PROPORTION p
a. A city government needs to determine the percent of its residents that are without health insurance
coverage. The city health department surveys 1600 city residents and finds that 248 of those included in
the survey lack (are without) health insurance. Construct and interpret a 95% confidence interval for the
true population proportion of all city residents who lack health insurance. Use a 95% confidence level.
population parameter: p = ____________________________________________________________
____________________________________________________________
random variable p′ = __________________________________________________________________
We are using sample data to estimate an unknown proportion for the whole population

HOW TO CALCULATE THE CONFIDENCE INTERVAL

Point Estimate = p ′ Confidence Level CL is area in the middle
Error Bound = (Critical Value)(Standard Error) Standard
Critical Value is Zα is the Z value that Error
p' q ' 2
EBP = Zα
2 n creates area of CL in the middle; Z ~N(0,1)
Use POSITIVE value of Z p' q '
Confidence Interval = Point Estimate + Error Bound n
Confidence Interval = p ′ + EBP invnorm (area to left , 0, 1)

Calculations and interpretation in context of the problem:

Page 3 of 8
EXAMPLE 1 Continued:
b. It has been estimated that for the state in which that city is located, approximately 20% of residents do
not have health insurance. Based on this confidence interval, can we conclude that the proportion of
all city residents who lack health insurance is lower than the proportion of state residents? Explain.

c. It has been estimated that nationally, approximately 16% of residents do not have health insurance.
Based on this confidence interval, can we conclude that the proportion of all city residents who
lack health insurance is lower than the proportion of U.S. residents? Explain.

d. What does it mean when we say the confidence level is 95% or when we say that we are "95% confident"?

 p' q ' 
The confidence interval for a proportion is based on the probability distribution N  p' , 
 n 

α p' q '  and invnorm  α p' q'  produces the same confidence interval.
Using invnorm  , p' ,
  1 − , p' , 
2 n   2 n 
This is because of the property we learned in chapter 6, that for the normal distribution the
probabilities found using the given distribution are always equal to the probabilities found using
the z-scores with the standard normal distribution Z~N(0,1)

invnorm(____ , _____ , _______ ) = _______ and invnorm(____ , _____, _______ ) = _______

Confidence Interval ( ________, ________ )

Page 4 of 8
 EXAMPLE 2: CONFIDENCE INTERVAL ESTIMATE for unknown POPULATION MEAN µ
when the POPULATION STANDARD DEVIATION σ is KNOWN
A soda bottling plant fills cans labeled to contain 12 ounces of soda. The filling machine varies and does
not fill each can with exactly 12 ounces. To determine if the filling machine needs adjustment, each day
the quality control manager measures the amount of soda per can for a random sample of 50 cans.
Experience shows that its filling machines have a known population standard deviation of 0.35 ounces.
In today's sample of 50 cans of soda, the sample average amount of soda per can is 12.1 ounces.
a. Construct and interpret a 90% confidence interval estimate for the true population average amount of
soda contained in all cans filled today at this bottling plant. Use a 90% confidence level.
X = __________________________________________________________________________________
population parameter: µ = ________________________________________________________________

random variable x = ____________________________________________________________________

We are using sample data to estimate an unknown mean (average) for the whole population
HOW TO CALCULATE THE CONFIDENCE INTERVAL for µ
When σ IS known, use the Standard normal distribution Z ~ N(0,1)
Point Estimate = x Confidence Level CL is area in the middle
Standard
Error Bound = (Critical Value)(Standard Error) Critical Value is Zα is the Z value that Error
EBM = Zα σ  2
  creates area of CL in the middle; Z ~N(0,1)
2  n
Use POSITIVE value of Z
σ
Confidence Interval = Point Estimate + Error Bound
Confidence Interval = x + EBM invnorm (area to left , 0, 1) n

Calculations and interpretation in context of the problem:

The confidence interval for µ when σ is known is based on the probability distribution 
N X ,
σ 

 n Page 5 of 8
EXAMPLE 3: CONFIDENCE INTERVAL ESTIMATE for unknown POPULATION MEAN µ
when the POPULATION STANDARD DEVIATION σ is NOT KNOWN
a. The speeds of 20 vehicles are observed by radar on a particular road.
For the vehicles in the sample, the average speed is 31.3 miles per hour with standard deviation 7.0 mph.
Construct and interpret a 98% confidence interval estimate of the true population average speed of all
vehicles traveling on this road. Use a 98% confidence level.
X = ________________________________________________________________________
population parameter: µ = ________________________________________________________________
random variable x = ____________________________________________________________________
We are using sample data to estimate an unknown mean (average) for the whole population
HOW TO CALCULATE THE CONFIDENCE INTERVAL for µ
When σ is NOT known, use the t distribution with degrees of freedom = sample size − 1 : t with df = n − 1)
Point Estimate = x Confidence Level CL is area in the middle
Error Bound = (Critical Value)(Standard Error) Critical Value tα is the t value that Standard
2
EBM = tα  s  Error
  creates an area of CL in the middle;
2  n  Use t distribution with df = n − 1 s
Confidence Interval = Point Estimate + Error Bound Use POSITIVE value of t n
Confidence Interval = x + EBM TI-84: tα = invT(area to left, df)
2

Calculations and interpretation in context of the problem:

___
b. In Example 3, suppose that you were not given the sample mean and sample standard deviation and instead
you were given a list of data for the speeds (in miles per hour) of the 20 vehicles.
19 19 22 24 25 27 28 37 35 30 37 36 39 40 43 30 31 36 33 35
How would you use the data to do this problem?

___
NOTE: Using the t-distribution requires that the underlying population of individual values is approximately
normally distributed. It is OK if this assumption is violated a little, but if the underlying population of individual
values has a distribution that differs too much from the normal distribution, then this confidence interval method
is not appropriate, and statisticians would use other techniques that we do not study in Math 10. Page 6 of 8
EXAMPLE 4: If we know the confidence interval: Working Backwards

The average nightly cost of hotel rooms for two resort areas are compared. Large random samples of hotel
room costs are collected for each city. The resulting confidence interval estimates are reported in a hotel
industry journal.

The 90% confidence interval estimate for the true population average nightly cost of all hotel rooms in
Surf City is $134 to $159 per night.
The 90% confidence interval estimate for the true population average nightly cost of all hotel rooms in
Ski Village is $123 to $141 per night.

a. Find the point estimate for the true average nightly cost of a hotel room in each city.
Which city has a higher point estimate?

b. Find the error bound for each city. Which city has a smaller margin of error?

c. Based on the confidence intervals only, would it be reasonable to conclude that the true average nightly
cost of a hotel rooms are different in Surf City and in Ski Village?

d. Would it be true that 90% of hotel rooms cost between $134 and $159 per night in Surf City and that
90% of hotel rooms cost between $123 and $141 per night in Ski Village?

Page 7 of 8
 Confidence Interval for a Proportion: Calculating the Sample Size needed in a Study
Given a desired confidence level and a desired margin of error, how large a sample is needed?

EBP = Zα p' q '

2 n
We know the error bound EBP that we want.
We know the confidence level CL we want, so we can find Zα corresponding to the desired CL.
2
We don't know p' or q' until we do the study, so we will assume for now that p' = q' = ½ = 0.5
Then we can substitute all these values into EBP = Zα p' q ' and solve for n.
2 n
2
p' q '  Zα 
Solving EBP = Zα for n gives . n =  2  p' q' .
2 n  EBP 
 
Sample Size Formula 2 The 0.25 appears in the formula because
Z 
to determine sample size n n =  α 2  (.25) we are assuming that p' = q' = ½ = 0.5
needed when estimating a  EBP 
  ALWAYS ROUND UP to the next higher integer
population proportion p

EXAMPLE 5: Finding the Sample Size:

a. Public opinion polls often do surveys with a 95% confidence level and a 3% margin of error.
Find the sample size needed to obtain this.
2
 Zα 
n=  2  (.25) =
 EBP 
 

b. Suppose a margin of error of 2% was wanted with a 95% confidence level. Find the sample size needed.
2
 Zα 
n=  2  (.25) =
 EBP 
 

c. Suppose a margin of error of 3% was wanted with a 90% confidence level. Find the sample size needed.

2
 Zα 
n=  2  (.25) =
 EBP 
 

d. Suppose a poll uses a sample size of n=100, and a confidence level of 95%.
Estimate the expected error bound using p' = q' = ½ = 0.5
p' q ' Note the actual error bound will differ after the
EBP = Zα study is done because we will know p' and q'
2 n and will no longer be estimating that p'=q'=0.5

e. Is the error bound in part d large or small compared to the examples in parts a, b, c?
Explain why this happened.

Page 8 of 8